Been thinking of my very old paper with Manuel on soft facts. The main inspiration there was Wittgenstein's remark that God needs to do maths in order to know something mathematical (for instance, how does pi expands). God cannot be a mere observer, any truth-monger is also an agent, in terms of my previous post. This goes well with the idea that scientists deal in construction, and construction is not something whose authority has to be found in their representational capacity. As Latour says somewhere in AIME, it is because they are good constructions that they are true. These are the idea behind soft facts: facts that are product of processes (or ententes). They cannot be reached unless you make the path that negotiation between agents require. In other words, years later, I guess that the idea of soft facts is tenable and recommendable. As it is formulated in the paper, its (implicit) account of agency - restricted to human - and God's - agents - tainted it.